ICE 1332/0715 Mobile Computing (Summer, 2008)

Transcription

1 ICE 1332/0715 Mobile Computing (Summer, 2008) TCP in Mobile Networks Prof. Chansu Yu Contents Vinton Cerf and Robert Kahn s paper in 1973 introduces TCP. Became standard in They received Turing Award in They are with MCI and NRI now. Physical layer issues Communication frequency Signal propagation Modulation and Demodulation Channel access issues Multiple access / Random access / Asynchronous / Bluetooth Capacity / Energy / Fairness / Directional System issues Embedded processor Low power design Network issues Location management Mobile IP / Cellular IP MANET routing / Clustering Multicast Interoperability Network reliability (TCP) Quality of service (QoS) TCP in Mobile IP Networks TCP in Mobile Ad Hoc Networks 2 1

2 Wireless Mobile Transport Layer Wireless environments are characterized by long latencies and frequent interruptions Problem: TCP has been optimized for wired networks Wired Network: When a packet is lost, it s typically a sign of congestion sender should slow down Wireless Network: When a packet gets lost, it could be due to Disconnects Long latencies slower transmission rates IP tunneling while node moves to new link What can it done? By whom? 3 TCP Issues Reducing the transmission rate is often the wrong response over wireless links. The sender should know the network it is transmitting over to make the right decision TCP issues in Mobile IP networks Handoff problem High transmission error rate TCP issues in MANET Impact of multiple-hop route Interplay with MAC (RTS, CTS, ACK) 4 2

3 Hand-off Problem Ramon Caceres, AT&T Bell Lab. Liviu Iftode, Princeton Univ. IEEE Journal on Selected Areas in Communication, 1995 (cited ~425). Hand-offs occur when a mobile host starts communicating with a new base station (in cellular wireless systems) 5 Hand-off Problem Hand-offs may result in temporary loss of route to MH with non-overlapping cells, it may take a while before the mobile host receives a beacon from the new BS While routes are being reestablished during handoff, MH and old BS may attempt to send packets to each other, resulting in loss of packets 6 3

4 Hand-off Problem Packet loss is mistaken as congestion Drops the transmission window size Slow start to restrict the windows growth rate Resets the retransmission timer to a backoff interval Thus, reduces the TCP throughput 7 Hand-off Problem During the long delay for a handoff to complete, a whole window worth of data may be lost After handoff is complete, acks are not received by the TCP sender Sender eventually times out, and retransmits If handoff still not complete, another timeout will occur Performance penalty Time wasted until timeout occurs Window shrunk after 8timeout 4

5 Mobility Scenarios 1. No handoffs 2. Overlapping Cells 3. 0-second rendezvous delay 4. 1-second rendezvous delay 9 0-second Rendezvous Delay : Beacon arrives as soon as cell boundary crossed Cell crossing + beacon arrives Handoff complete Routes updated Retransmission timeout Last timed transmit RTO=1.0 can be improved sec Packet loss 10 Idle sender 5

6 Mobility Scenarios 1. No handoffs 2. Overlapping Cells 3. 0-second rendezvous delay 4. 1-second rendezvous delay 11 1-second Rendezvous Delay : Beacon arrives 1 second after cell boundary crossed Last timed transmit Cell crossing RTO=1.0 Timeout 1 Beacon arrives Handoff complete RTO=2.0 Retransmission timeout sec can be improved Packet loss 12 Idle sender 6

7 TCP Performance (1-sec delay) 4 60 ) s te 3 y b M r( e b m 2 u n e c n e u q e 1 S s ) 40 y te b ( K e iz s w o d 20 in W Time(s econds ) Time(seconds ) cell crossing <Behavior of TCP sequence number> <Behavior of TCP congestion window> 13 TCP Performance A v e ra ge th rou gh put (Kbit/ s e c ) MH switches cells every 8 seconds Throughput dropped significantly in the presence of motion Degradation in overlapping cells is due to encapsulation and forwarding delay during handoff No h a n doffs O v e rla ppin g c ells 0-s ec on d 1-s e c on d re n dezv ou s ren de zv ou s de lay de la y 14 Additional degradation in cases 3 and 4 due to packet loss and idle time at sender 7

8 Approaches TCP issues in Mobile IP networks Handoff problem - Fast retransmission High transmission error rate - Split connection TCP issues in MANET Impact of multiple-hop route Interplay with MAC 15 Handoff Problem - Fast Retransmission When the packet loss is due to handoff, who can make the right decision? And which action can be taken? When MH is the TCP receiver: after handoff is complete, it sends 3 dupacks to the sender this triggers fast retransmit at the sender instead of dupacks, a special notification could also be sent When MH is the TCP sender: invoke fast retransmit after completion of handoff 16 8

9 0-second Rendezvous Delay Improvement using Fast Retransmit Cell crossing + Beacon arrives Handoff completes Retransmission timeout Fast retransmission RTO=1.0 Does not occur!!! Last timed transmission Time(seconds) Fast retransmission after a handoff with a 0-second rendezvous delay 17 1-second Rendezvous Delay Improvement using Fast Retransmit Cell crossing Retransmission timeout 1 Beacon arrives Handoff complete Retransmission timeout 2 Last timed transmission RTO=1.0 Fast retransmission RTO=2.0 Does not occur!!! Time(seconds) Handoff latency and related packet losses with a 1-second rendezvous delay 18 9

10 TCP Performance Improvement TCP 1983 TCP Tahoe 1988 (Fast retransmit, by Van Jacobson, LBL) TCP Reno 1990 (Fast recovery) TCP Vegas 1994 (Congestion prediction by U.Arozona) A v e r a ge t h ro u gh p u t (Kbit / s e c ) 100% % % % 88% % 1100 No change in the first two cases as expected Improvement for nonoverlapping cells Some degradation still remains fast retransmit reduces congestion window No h a nd o f fs Ov e rla p p ing c e lls 0-s e c o n d re n de z v o u s d e la y 1-s e c on d r en d e zv o u s d ela y Do we need to change TCP software? 19 Approaches TCP issues in Mobile IP networks Handoff problem - Fast retransmission High transmission error rate - Split connection TCP issues in MANET Impact of multiple-hop route Interplay with MAC 20 10

11 High Transmission Error Rate - Split Connection Approach Ho can we address the problem of high error rate over wireless links? End-to-end TCP connection is broken into one connection on the wired part of the route and one over wireless part of the route A single TCP connection split into two TCP connections FH-MH = FH-BS + BS-MH Acks are intercepted and managed at BS FH BS MH Fixed Host Base Station Mobile Host 21 Split Connection Approach Per-TCP connection state TCP connection TCP connection application transport application transport rxmt application transport network network network link link link physical physical physical FH BS 22 wireless MH 11

12 Split Connection Approach: What happen if move mobile host FA 2 Internet standard TCP mobile host wireless TCP FA 1 23 Split Connection Approach: State Migration mobile host wireless TCP FA 2 standard TCP socket migration and state transfer Internet FA

13 Split Connection Approach BS terminates the standard TCP connection acting as a proxy Old BS (FA) must migrate buffered packets (already acknowledged to FH) as well as socket of the proxy to new BS The socket contains the current state of the TCP connection Sequence number, addresses, port number Last packet transmitted to MH Last packet acknowledged by MH Next expected acknowledgement and expected number of duplicated acknowledgements Round-trip time estimate of wireless link 25 Split Connection Approach: Variations Indirect TCP FH - BS connection : Standard TCP BS - MH connection : Standard TCP Selective Repeat Protocol (SRP) FH - BS connection : standard TCP BS - MH connection : selective repeat protocol on top of UDP Asymmetric transport protocol (Mobile-TCP) Low overhead protocol at wireless hosts such as header compression, simpler flow control, No congestion control Mobile-End Transport Protocol BS-MH link can use any arbitrary protocol optimized for wireless link 26 13

14 Split Connection Approach : Advantages BS-MH connection can be optimized independent of FH-BS connection Different flow / error control on the two connections Local recovery of errors Faster recovery due to relatively shorter RTT on wireless link Good performance achievable using appropriate BS-MH protocol Standard TCP on BS-MH performs poorly when multiple packet losses occur per window (timeouts can occur on the BS-MH connection, stalling during the timeout interval) Selective acks improve performance for such cases 27 Split Connection Approach : Disadvantages End-to-end semantics violated ack may be delivered to sender, before data delivered to the receiver FH BS MH

15 Split Connection Approach : Disadvantages BS retains hard state BS failure can result in loss of data (unreliability) If BS fails, packet 40 will be lost Since it is ack d to sender, the sender does not buffer 40 Hand-off latency increases due to state transfer Data that has been ack d to sender, must be moved to new base station Buffer space needed at BS for each TCP connection BS buffers tend to get full, when wireless link slower (one window worth of data on wired connection could be stored at the base station, for each split connection) Extra copying of data at BS copying from FH-BS socket buffer to BS-MH socket buffer (at TCP layer) increases end-to-end latency 29 Snoop Protocol: TCP-Aware Link Layer Tries to restore the end-to-end semantics of TCP: Foreign agent is not allowed to send acknowledgements Instead, it just snoops on packets and tries to help along correspondent host Internet mobile host local retransmission Foreign agent snooping of ACKs end-to-end TCP connection 30 buffering of data 15

16 Snoop Protocol: TCP-Aware Link Layer Per TCP-connection state TCP connection application application application transport transport transport network link network link rxmt network link physical physical physical FH BS wireless MH 31 Approaches TCP issues in Mobile IP networks Handoff problem - Fast retransmission High transmission error rate - Split connection TCP issues in MANET Impact of multiple-hop route Interplay with MAC 32 16

17 TCP in MANET G. Holland and N. H. Vaidya, ACM Mobicom, 1999 (UIUC). Connections over multiple hops are at a disadvantage compared to shorter connections, because they have to contend for wireless access at each hop Number of hops TCP Throughtput (Kbps) TCP Throughput using 2 Mbps MAC Throughput Degradations with Increasing Number of Hops Packet transmission can occur on at most one hop among three consecutive hops Increasing the number of hops from 1 to 2, 3 results in increased delay, and decreased throughput Increasing number of hops beyond 3 allows simultaneous transmissions on more than one link. however, degradation continues due to contention between TCP Data and Acks traveling in opposite directions When number of hops is large enough, the throughput stabilizes due to effective pipelining 34 17

18 Why Does Throughput Degrade? mobility causes link breakage, resulting in route failure TCP sender times out. (RTO) Route is repaired TCP sender times out. Resumes sending No throughput No throughput despite route repair TCP data and acks en route discarded 35 Larger route repair delays especially harmful (Route discovery) Interplay with MAC S. Xu and T. Saadawi. IEEE Communications Magazine, 2001 (CUNY). Does MAC function well in multi-hop networks? If not, why? And how to alleviate the problem? Experiment Methods: A Static String Topology TCP as Transport Layer Protocol Issues Instability Problem Unfairness Problem 36 18

19 Simulation Environment Simulator: ns-2 MAC Layer: IEEE DCF Transport Layer: TCP connections carrying large files Network Environment A Static String Network Topology Interfering range is a little more than two times of the communication range Interfering Range Communication Range Instability Problem Experiment Setup Source Destination A single TCP connection, with node 1 as the source and node 5 as the destination. Three sets of experiments with Maximum Window Size (window_) 32, 8, and 4 respectively

20 Instability Problem Experiment Result When window_=32 or 8, serious oscillation of throughput is observed. When window_4, throughput is stable. 39 Instability Problem Trace Analysis(1) Interfering Range of Node 2 RTS Data CTS Ack RTS is sent but not received RTS is sent and received but CTS can t be sent 40 20

21 Instability Problem Summary Collision and exposed terminal problem prevent node 2 from receiving RTS from or sending CTS to node 1. The random backoff, big data packet, and sending back-to-back packets worsen the above problems. When window_ = 4, the chance to send back a CTS is greatly increased, so the throughput becomes stable. After node 1 fails seven times to receive CTS, node 1 believes there is a route failure and starts a route discovery. Before a route is available, node 1 cannot send out a data packet. This period usually is long enough to cause a timeout at the TCP sender. For TCP, timeout triggers Slow Start, which significantly reduces the 41throughput. Unfairness Problem Experiment Setup Second Session First Session Source Dest Dest Source In the first session, data flow from 6 to 4. In the second session, data flow from 2 to 3. The first session starts at 10.0s. The second session starts at 30.0s

22 Unfairness Problem Experiment Result(1) From 10s to 30s (only the first session exists) The first session has a throughput of about 450kbps After 30s (with two sessions) The second session has a throughput of about 900kbps while the first has zero throughput 43 Unfairness Problem Experiment Result(2) The first session never succeeds to send out packet with sequence number What happened to the first session around 30s? No ACKs and TCP drops One packet done No ACK for the next one enlarged 44 22

23 Unfairness Problem Trace Analysis(1) Interfering Range of Node 5 Data RTS Data Ack CTS No Route Interfering Range of Node 4 What happen? What is this? What happen? 45 Unfairness Problem Trace Analysis(2) Interfering Range of Node 5 Data RTS Data Ack CTS Interfering Range of Node 4 No Route RTS is sent but not received 46 23

24 Summary: TCP traffic in multihop networks Instability Problem: TCP throughput fluctuates Unfairness problem One-hop connection has more chances to transmit data because the interval between packet transmission is shorter than that of the multi-hop TCP connections Random back-off is actually advantageous to the last succeeding host. Conclusions: IEEE does not work well in multi-hop wireless networks. Conventional transport protocol does not work well in multi-hop wireless networks